I have a rudimentary patch which appears to enable
Unicode support, allowing you to specify an encoding
for Unicode strings passed into the database adapter.
I'm not sure if doing s.encode() on strings passed into
the string_literal() function is a really good idea
(maybe you can get away with evil things that won't be
correctly escaped now?) but it appears to work.
It also doesn't test that the underlying database is
happy with Unicode -- something I think only turned up
in 4.1 onwards?